4034def16c7680bafcb6364298112f7ff459c819
1#include "cache.h"
2#include "tag.h"
3#include "commit.h"
4#include "commit-graph.h"
5#include "repository.h"
6#include "object-store.h"
7#include "pkt-line.h"
8#include "utf8.h"
9#include "diff.h"
10#include "revision.h"
11#include "notes.h"
12#include "alloc.h"
13#include "gpg-interface.h"
14#include "mergesort.h"
15#include "commit-slab.h"
16#include "prio-queue.h"
17#include "sha1-lookup.h"
18#include "wt-status.h"
19#include "advice.h"
20
21static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
22
23int save_commit_buffer = 1;
24
25const char *commit_type = "commit";
26
27struct commit *lookup_commit_reference_gently(struct repository *r,
28 const struct object_id *oid, int quiet)
29{
30 struct object *obj = deref_tag(r,
31 parse_object(r, oid),
32 NULL, 0);
33
34 if (!obj)
35 return NULL;
36 return object_as_type(r, obj, OBJ_COMMIT, quiet);
37}
38
39struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
40{
41 return lookup_commit_reference_gently(r, oid, 0);
42}
43
44struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
45{
46 struct commit *c = lookup_commit_reference(the_repository, oid);
47 if (!c)
48 die(_("could not parse %s"), ref_name);
49 if (!oideq(oid, &c->object.oid)) {
50 warning(_("%s %s is not a commit!"),
51 ref_name, oid_to_hex(oid));
52 }
53 return c;
54}
55
56struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
57{
58 struct object *obj = lookup_object(r, oid->hash);
59 if (!obj)
60 return create_object(r, oid->hash,
61 alloc_commit_node(r));
62 return object_as_type(r, obj, OBJ_COMMIT, 0);
63}
64
65struct commit *lookup_commit_reference_by_name(const char *name)
66{
67 struct object_id oid;
68 struct commit *commit;
69
70 if (get_oid_committish(name, &oid))
71 return NULL;
72 commit = lookup_commit_reference(the_repository, &oid);
73 if (parse_commit(commit))
74 return NULL;
75 return commit;
76}
77
78static timestamp_t parse_commit_date(const char *buf, const char *tail)
79{
80 const char *dateptr;
81
82 if (buf + 6 >= tail)
83 return 0;
84 if (memcmp(buf, "author", 6))
85 return 0;
86 while (buf < tail && *buf++ != '\n')
87 /* nada */;
88 if (buf + 9 >= tail)
89 return 0;
90 if (memcmp(buf, "committer", 9))
91 return 0;
92 while (buf < tail && *buf++ != '>')
93 /* nada */;
94 if (buf >= tail)
95 return 0;
96 dateptr = buf;
97 while (buf < tail && *buf++ != '\n')
98 /* nada */;
99 if (buf >= tail)
100 return 0;
101 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
102 return parse_timestamp(dateptr, NULL, 10);
103}
104
105static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
106{
107 struct commit_graft **commit_graft_table = table;
108 return commit_graft_table[index]->oid.hash;
109}
110
111static int commit_graft_pos(struct repository *r, const unsigned char *sha1)
112{
113 return sha1_pos(sha1, r->parsed_objects->grafts,
114 r->parsed_objects->grafts_nr,
115 commit_graft_sha1_access);
116}
117
118int register_commit_graft(struct repository *r, struct commit_graft *graft,
119 int ignore_dups)
120{
121 int pos = commit_graft_pos(r, graft->oid.hash);
122
123 if (0 <= pos) {
124 if (ignore_dups)
125 free(graft);
126 else {
127 free(r->parsed_objects->grafts[pos]);
128 r->parsed_objects->grafts[pos] = graft;
129 }
130 return 1;
131 }
132 pos = -pos - 1;
133 ALLOC_GROW(r->parsed_objects->grafts,
134 r->parsed_objects->grafts_nr + 1,
135 r->parsed_objects->grafts_alloc);
136 r->parsed_objects->grafts_nr++;
137 if (pos < r->parsed_objects->grafts_nr)
138 memmove(r->parsed_objects->grafts + pos + 1,
139 r->parsed_objects->grafts + pos,
140 (r->parsed_objects->grafts_nr - pos - 1) *
141 sizeof(*r->parsed_objects->grafts));
142 r->parsed_objects->grafts[pos] = graft;
143 return 0;
144}
145
146struct commit_graft *read_graft_line(struct strbuf *line)
147{
148 /* The format is just "Commit Parent1 Parent2 ...\n" */
149 int i, phase;
150 const char *tail = NULL;
151 struct commit_graft *graft = NULL;
152 struct object_id dummy_oid, *oid;
153
154 strbuf_rtrim(line);
155 if (!line->len || line->buf[0] == '#')
156 return NULL;
157 /*
158 * phase 0 verifies line, counts hashes in line and allocates graft
159 * phase 1 fills graft
160 */
161 for (phase = 0; phase < 2; phase++) {
162 oid = graft ? &graft->oid : &dummy_oid;
163 if (parse_oid_hex(line->buf, oid, &tail))
164 goto bad_graft_data;
165 for (i = 0; *tail != '\0'; i++) {
166 oid = graft ? &graft->parent[i] : &dummy_oid;
167 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
168 goto bad_graft_data;
169 }
170 if (!graft) {
171 graft = xmalloc(st_add(sizeof(*graft),
172 st_mult(sizeof(struct object_id), i)));
173 graft->nr_parent = i;
174 }
175 }
176 return graft;
177
178bad_graft_data:
179 error("bad graft data: %s", line->buf);
180 assert(!graft);
181 return NULL;
182}
183
184static int read_graft_file(struct repository *r, const char *graft_file)
185{
186 FILE *fp = fopen_or_warn(graft_file, "r");
187 struct strbuf buf = STRBUF_INIT;
188 if (!fp)
189 return -1;
190 if (advice_graft_file_deprecated)
191 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
192 "and will be removed in a future Git version.\n"
193 "\n"
194 "Please use \"git replace --convert-graft-file\"\n"
195 "to convert the grafts into replace refs.\n"
196 "\n"
197 "Turn this message off by running\n"
198 "\"git config advice.graftFileDeprecated false\""));
199 while (!strbuf_getwholeline(&buf, fp, '\n')) {
200 /* The format is just "Commit Parent1 Parent2 ...\n" */
201 struct commit_graft *graft = read_graft_line(&buf);
202 if (!graft)
203 continue;
204 if (register_commit_graft(r, graft, 1))
205 error("duplicate graft data: %s", buf.buf);
206 }
207 fclose(fp);
208 strbuf_release(&buf);
209 return 0;
210}
211
212void prepare_commit_graft(struct repository *r)
213{
214 char *graft_file;
215
216 if (r->parsed_objects->commit_graft_prepared)
217 return;
218 if (!startup_info->have_repository)
219 return;
220
221 graft_file = get_graft_file(r);
222 read_graft_file(r, graft_file);
223 /* make sure shallows are read */
224 is_repository_shallow(r);
225 r->parsed_objects->commit_graft_prepared = 1;
226}
227
228struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
229{
230 int pos;
231 prepare_commit_graft(r);
232 pos = commit_graft_pos(r, oid->hash);
233 if (pos < 0)
234 return NULL;
235 return r->parsed_objects->grafts[pos];
236}
237
238int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
239{
240 int i, ret;
241 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
242 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
243 return ret;
244}
245
246int unregister_shallow(const struct object_id *oid)
247{
248 int pos = commit_graft_pos(the_repository, oid->hash);
249 if (pos < 0)
250 return -1;
251 if (pos + 1 < the_repository->parsed_objects->grafts_nr)
252 MOVE_ARRAY(the_repository->parsed_objects->grafts + pos,
253 the_repository->parsed_objects->grafts + pos + 1,
254 the_repository->parsed_objects->grafts_nr - pos - 1);
255 the_repository->parsed_objects->grafts_nr--;
256 return 0;
257}
258
259struct commit_buffer {
260 void *buffer;
261 unsigned long size;
262};
263define_commit_slab(buffer_slab, struct commit_buffer);
264
265struct buffer_slab *allocate_commit_buffer_slab(void)
266{
267 struct buffer_slab *bs = xmalloc(sizeof(*bs));
268 init_buffer_slab(bs);
269 return bs;
270}
271
272void free_commit_buffer_slab(struct buffer_slab *bs)
273{
274 clear_buffer_slab(bs);
275 free(bs);
276}
277
278void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
279{
280 struct commit_buffer *v = buffer_slab_at(
281 r->parsed_objects->buffer_slab, commit);
282 v->buffer = buffer;
283 v->size = size;
284}
285
286const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
287{
288 struct commit_buffer *v = buffer_slab_peek(
289 r->parsed_objects->buffer_slab, commit);
290 if (!v) {
291 if (sizep)
292 *sizep = 0;
293 return NULL;
294 }
295 if (sizep)
296 *sizep = v->size;
297 return v->buffer;
298}
299
300const void *repo_get_commit_buffer(struct repository *r,
301 const struct commit *commit,
302 unsigned long *sizep)
303{
304 const void *ret = get_cached_commit_buffer(r, commit, sizep);
305 if (!ret) {
306 enum object_type type;
307 unsigned long size;
308 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
309 if (!ret)
310 die("cannot read commit object %s",
311 oid_to_hex(&commit->object.oid));
312 if (type != OBJ_COMMIT)
313 die("expected commit for %s, got %s",
314 oid_to_hex(&commit->object.oid), type_name(type));
315 if (sizep)
316 *sizep = size;
317 }
318 return ret;
319}
320
321void unuse_commit_buffer(const struct commit *commit, const void *buffer)
322{
323 struct commit_buffer *v = buffer_slab_peek(
324 the_repository->parsed_objects->buffer_slab, commit);
325 if (!(v && v->buffer == buffer))
326 free((void *)buffer);
327}
328
329void free_commit_buffer(struct commit *commit)
330{
331 struct commit_buffer *v = buffer_slab_peek(
332 the_repository->parsed_objects->buffer_slab, commit);
333 if (v) {
334 FREE_AND_NULL(v->buffer);
335 v->size = 0;
336 }
337}
338
339struct tree *get_commit_tree(const struct commit *commit)
340{
341 if (commit->maybe_tree || !commit->object.parsed)
342 return commit->maybe_tree;
343
344 if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
345 BUG("commit has NULL tree, but was not loaded from commit-graph");
346
347 return get_commit_tree_in_graph(the_repository, commit);
348}
349
350struct object_id *get_commit_tree_oid(const struct commit *commit)
351{
352 return &get_commit_tree(commit)->object.oid;
353}
354
355void release_commit_memory(struct commit *c)
356{
357 c->maybe_tree = NULL;
358 c->index = 0;
359 free_commit_buffer(c);
360 free_commit_list(c->parents);
361 /* TODO: what about commit->util? */
362
363 c->object.parsed = 0;
364}
365
366const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
367{
368 struct commit_buffer *v = buffer_slab_peek(
369 the_repository->parsed_objects->buffer_slab, commit);
370 void *ret;
371
372 if (!v) {
373 if (sizep)
374 *sizep = 0;
375 return NULL;
376 }
377 ret = v->buffer;
378 if (sizep)
379 *sizep = v->size;
380
381 v->buffer = NULL;
382 v->size = 0;
383 return ret;
384}
385
386int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
387{
388 const char *tail = buffer;
389 const char *bufptr = buffer;
390 struct object_id parent;
391 struct commit_list **pptr;
392 struct commit_graft *graft;
393 const int tree_entry_len = the_hash_algo->hexsz + 5;
394 const int parent_entry_len = the_hash_algo->hexsz + 7;
395
396 if (item->object.parsed)
397 return 0;
398 item->object.parsed = 1;
399 tail += size;
400 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
401 bufptr[tree_entry_len] != '\n')
402 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
403 if (get_oid_hex(bufptr + 5, &parent) < 0)
404 return error("bad tree pointer in commit %s",
405 oid_to_hex(&item->object.oid));
406 item->maybe_tree = lookup_tree(r, &parent);
407 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
408 pptr = &item->parents;
409
410 graft = lookup_commit_graft(r, &item->object.oid);
411 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
412 struct commit *new_parent;
413
414 if (tail <= bufptr + parent_entry_len + 1 ||
415 get_oid_hex(bufptr + 7, &parent) ||
416 bufptr[parent_entry_len] != '\n')
417 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
418 bufptr += parent_entry_len + 1;
419 /*
420 * The clone is shallow if nr_parent < 0, and we must
421 * not traverse its real parents even when we unhide them.
422 */
423 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
424 continue;
425 new_parent = lookup_commit(r, &parent);
426 if (new_parent)
427 pptr = &commit_list_insert(new_parent, pptr)->next;
428 }
429 if (graft) {
430 int i;
431 struct commit *new_parent;
432 for (i = 0; i < graft->nr_parent; i++) {
433 new_parent = lookup_commit(r,
434 &graft->parent[i]);
435 if (!new_parent)
436 continue;
437 pptr = &commit_list_insert(new_parent, pptr)->next;
438 }
439 }
440 item->date = parse_commit_date(bufptr, tail);
441
442 if (check_graph)
443 load_commit_graph_info(the_repository, item);
444
445 return 0;
446}
447
448int repo_parse_commit_internal(struct repository *r,
449 struct commit *item,
450 int quiet_on_missing,
451 int use_commit_graph)
452{
453 enum object_type type;
454 void *buffer;
455 unsigned long size;
456 int ret;
457
458 if (!item)
459 return -1;
460 if (item->object.parsed)
461 return 0;
462 if (use_commit_graph && parse_commit_in_graph(r, item))
463 return 0;
464 buffer = repo_read_object_file(r, &item->object.oid, &type, &size);
465 if (!buffer)
466 return quiet_on_missing ? -1 :
467 error("Could not read %s",
468 oid_to_hex(&item->object.oid));
469 if (type != OBJ_COMMIT) {
470 free(buffer);
471 return error("Object %s not a commit",
472 oid_to_hex(&item->object.oid));
473 }
474
475 ret = parse_commit_buffer(r, item, buffer, size, 0);
476 if (save_commit_buffer && !ret) {
477 set_commit_buffer(r, item, buffer, size);
478 return 0;
479 }
480 free(buffer);
481 return ret;
482}
483
484int repo_parse_commit_gently(struct repository *r,
485 struct commit *item, int quiet_on_missing)
486{
487 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
488}
489
490void parse_commit_or_die(struct commit *item)
491{
492 if (parse_commit(item))
493 die("unable to parse commit %s",
494 item ? oid_to_hex(&item->object.oid) : "(null)");
495}
496
497int find_commit_subject(const char *commit_buffer, const char **subject)
498{
499 const char *eol;
500 const char *p = commit_buffer;
501
502 while (*p && (*p != '\n' || p[1] != '\n'))
503 p++;
504 if (*p) {
505 p = skip_blank_lines(p + 2);
506 eol = strchrnul(p, '\n');
507 } else
508 eol = p;
509
510 *subject = p;
511
512 return eol - p;
513}
514
515struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
516{
517 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
518 new_list->item = item;
519 new_list->next = *list_p;
520 *list_p = new_list;
521 return new_list;
522}
523
524unsigned commit_list_count(const struct commit_list *l)
525{
526 unsigned c = 0;
527 for (; l; l = l->next )
528 c++;
529 return c;
530}
531
532struct commit_list *copy_commit_list(struct commit_list *list)
533{
534 struct commit_list *head = NULL;
535 struct commit_list **pp = &head;
536 while (list) {
537 pp = commit_list_append(list->item, pp);
538 list = list->next;
539 }
540 return head;
541}
542
543void free_commit_list(struct commit_list *list)
544{
545 while (list)
546 pop_commit(&list);
547}
548
549struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
550{
551 struct commit_list **pp = list;
552 struct commit_list *p;
553 while ((p = *pp) != NULL) {
554 if (p->item->date < item->date) {
555 break;
556 }
557 pp = &p->next;
558 }
559 return commit_list_insert(item, pp);
560}
561
562static int commit_list_compare_by_date(const void *a, const void *b)
563{
564 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
565 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
566 if (a_date < b_date)
567 return 1;
568 if (a_date > b_date)
569 return -1;
570 return 0;
571}
572
573static void *commit_list_get_next(const void *a)
574{
575 return ((const struct commit_list *)a)->next;
576}
577
578static void commit_list_set_next(void *a, void *next)
579{
580 ((struct commit_list *)a)->next = next;
581}
582
583void commit_list_sort_by_date(struct commit_list **list)
584{
585 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
586 commit_list_compare_by_date);
587}
588
589struct commit *pop_most_recent_commit(struct commit_list **list,
590 unsigned int mark)
591{
592 struct commit *ret = pop_commit(list);
593 struct commit_list *parents = ret->parents;
594
595 while (parents) {
596 struct commit *commit = parents->item;
597 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
598 commit->object.flags |= mark;
599 commit_list_insert_by_date(commit, list);
600 }
601 parents = parents->next;
602 }
603 return ret;
604}
605
606static void clear_commit_marks_1(struct commit_list **plist,
607 struct commit *commit, unsigned int mark)
608{
609 while (commit) {
610 struct commit_list *parents;
611
612 if (!(mark & commit->object.flags))
613 return;
614
615 commit->object.flags &= ~mark;
616
617 parents = commit->parents;
618 if (!parents)
619 return;
620
621 while ((parents = parents->next))
622 commit_list_insert(parents->item, plist);
623
624 commit = commit->parents->item;
625 }
626}
627
628void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
629{
630 struct commit_list *list = NULL;
631
632 while (nr--) {
633 clear_commit_marks_1(&list, *commit, mark);
634 commit++;
635 }
636 while (list)
637 clear_commit_marks_1(&list, pop_commit(&list), mark);
638}
639
640void clear_commit_marks(struct commit *commit, unsigned int mark)
641{
642 clear_commit_marks_many(1, &commit, mark);
643}
644
645struct commit *pop_commit(struct commit_list **stack)
646{
647 struct commit_list *top = *stack;
648 struct commit *item = top ? top->item : NULL;
649
650 if (top) {
651 *stack = top->next;
652 free(top);
653 }
654 return item;
655}
656
657/*
658 * Topological sort support
659 */
660
661/* count number of children that have not been emitted */
662define_commit_slab(indegree_slab, int);
663
664/* record author-date for each commit object */
665define_commit_slab(author_date_slab, timestamp_t);
666
667static void record_author_date(struct author_date_slab *author_date,
668 struct commit *commit)
669{
670 const char *buffer = get_commit_buffer(commit, NULL);
671 struct ident_split ident;
672 const char *ident_line;
673 size_t ident_len;
674 char *date_end;
675 timestamp_t date;
676
677 ident_line = find_commit_header(buffer, "author", &ident_len);
678 if (!ident_line)
679 goto fail_exit; /* no author line */
680 if (split_ident_line(&ident, ident_line, ident_len) ||
681 !ident.date_begin || !ident.date_end)
682 goto fail_exit; /* malformed "author" line */
683
684 date = parse_timestamp(ident.date_begin, &date_end, 10);
685 if (date_end != ident.date_end)
686 goto fail_exit; /* malformed date */
687 *(author_date_slab_at(author_date, commit)) = date;
688
689fail_exit:
690 unuse_commit_buffer(commit, buffer);
691}
692
693static int compare_commits_by_author_date(const void *a_, const void *b_,
694 void *cb_data)
695{
696 const struct commit *a = a_, *b = b_;
697 struct author_date_slab *author_date = cb_data;
698 timestamp_t a_date = *(author_date_slab_at(author_date, a));
699 timestamp_t b_date = *(author_date_slab_at(author_date, b));
700
701 /* newer commits with larger date first */
702 if (a_date < b_date)
703 return 1;
704 else if (a_date > b_date)
705 return -1;
706 return 0;
707}
708
709int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
710{
711 const struct commit *a = a_, *b = b_;
712
713 /* newer commits first */
714 if (a->generation < b->generation)
715 return 1;
716 else if (a->generation > b->generation)
717 return -1;
718
719 /* use date as a heuristic when generations are equal */
720 if (a->date < b->date)
721 return 1;
722 else if (a->date > b->date)
723 return -1;
724 return 0;
725}
726
727int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
728{
729 const struct commit *a = a_, *b = b_;
730 /* newer commits with larger date first */
731 if (a->date < b->date)
732 return 1;
733 else if (a->date > b->date)
734 return -1;
735 return 0;
736}
737
738/*
739 * Performs an in-place topological sort on the list supplied.
740 */
741void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
742{
743 struct commit_list *next, *orig = *list;
744 struct commit_list **pptr;
745 struct indegree_slab indegree;
746 struct prio_queue queue;
747 struct commit *commit;
748 struct author_date_slab author_date;
749
750 if (!orig)
751 return;
752 *list = NULL;
753
754 init_indegree_slab(&indegree);
755 memset(&queue, '\0', sizeof(queue));
756
757 switch (sort_order) {
758 default: /* REV_SORT_IN_GRAPH_ORDER */
759 queue.compare = NULL;
760 break;
761 case REV_SORT_BY_COMMIT_DATE:
762 queue.compare = compare_commits_by_commit_date;
763 break;
764 case REV_SORT_BY_AUTHOR_DATE:
765 init_author_date_slab(&author_date);
766 queue.compare = compare_commits_by_author_date;
767 queue.cb_data = &author_date;
768 break;
769 }
770
771 /* Mark them and clear the indegree */
772 for (next = orig; next; next = next->next) {
773 struct commit *commit = next->item;
774 *(indegree_slab_at(&indegree, commit)) = 1;
775 /* also record the author dates, if needed */
776 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
777 record_author_date(&author_date, commit);
778 }
779
780 /* update the indegree */
781 for (next = orig; next; next = next->next) {
782 struct commit_list *parents = next->item->parents;
783 while (parents) {
784 struct commit *parent = parents->item;
785 int *pi = indegree_slab_at(&indegree, parent);
786
787 if (*pi)
788 (*pi)++;
789 parents = parents->next;
790 }
791 }
792
793 /*
794 * find the tips
795 *
796 * tips are nodes not reachable from any other node in the list
797 *
798 * the tips serve as a starting set for the work queue.
799 */
800 for (next = orig; next; next = next->next) {
801 struct commit *commit = next->item;
802
803 if (*(indegree_slab_at(&indegree, commit)) == 1)
804 prio_queue_put(&queue, commit);
805 }
806
807 /*
808 * This is unfortunate; the initial tips need to be shown
809 * in the order given from the revision traversal machinery.
810 */
811 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
812 prio_queue_reverse(&queue);
813
814 /* We no longer need the commit list */
815 free_commit_list(orig);
816
817 pptr = list;
818 *list = NULL;
819 while ((commit = prio_queue_get(&queue)) != NULL) {
820 struct commit_list *parents;
821
822 for (parents = commit->parents; parents ; parents = parents->next) {
823 struct commit *parent = parents->item;
824 int *pi = indegree_slab_at(&indegree, parent);
825
826 if (!*pi)
827 continue;
828
829 /*
830 * parents are only enqueued for emission
831 * when all their children have been emitted thereby
832 * guaranteeing topological order.
833 */
834 if (--(*pi) == 1)
835 prio_queue_put(&queue, parent);
836 }
837 /*
838 * all children of commit have already been
839 * emitted. we can emit it now.
840 */
841 *(indegree_slab_at(&indegree, commit)) = 0;
842
843 pptr = &commit_list_insert(commit, pptr)->next;
844 }
845
846 clear_indegree_slab(&indegree);
847 clear_prio_queue(&queue);
848 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
849 clear_author_date_slab(&author_date);
850}
851
852static const char gpg_sig_header[] = "gpgsig";
853static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
854
855static int do_sign_commit(struct strbuf *buf, const char *keyid)
856{
857 struct strbuf sig = STRBUF_INIT;
858 int inspos, copypos;
859 const char *eoh;
860
861 /* find the end of the header */
862 eoh = strstr(buf->buf, "\n\n");
863 if (!eoh)
864 inspos = buf->len;
865 else
866 inspos = eoh - buf->buf + 1;
867
868 if (!keyid || !*keyid)
869 keyid = get_signing_key();
870 if (sign_buffer(buf, &sig, keyid)) {
871 strbuf_release(&sig);
872 return -1;
873 }
874
875 for (copypos = 0; sig.buf[copypos]; ) {
876 const char *bol = sig.buf + copypos;
877 const char *eol = strchrnul(bol, '\n');
878 int len = (eol - bol) + !!*eol;
879
880 if (!copypos) {
881 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
882 inspos += gpg_sig_header_len;
883 }
884 strbuf_insert(buf, inspos++, " ", 1);
885 strbuf_insert(buf, inspos, bol, len);
886 inspos += len;
887 copypos += len;
888 }
889 strbuf_release(&sig);
890 return 0;
891}
892
893int parse_signed_commit(const struct commit *commit,
894 struct strbuf *payload, struct strbuf *signature)
895{
896
897 unsigned long size;
898 const char *buffer = get_commit_buffer(commit, &size);
899 int in_signature, saw_signature = -1;
900 const char *line, *tail;
901
902 line = buffer;
903 tail = buffer + size;
904 in_signature = 0;
905 saw_signature = 0;
906 while (line < tail) {
907 const char *sig = NULL;
908 const char *next = memchr(line, '\n', tail - line);
909
910 next = next ? next + 1 : tail;
911 if (in_signature && line[0] == ' ')
912 sig = line + 1;
913 else if (starts_with(line, gpg_sig_header) &&
914 line[gpg_sig_header_len] == ' ')
915 sig = line + gpg_sig_header_len + 1;
916 if (sig) {
917 strbuf_add(signature, sig, next - sig);
918 saw_signature = 1;
919 in_signature = 1;
920 } else {
921 if (*line == '\n')
922 /* dump the whole remainder of the buffer */
923 next = tail;
924 strbuf_add(payload, line, next - line);
925 in_signature = 0;
926 }
927 line = next;
928 }
929 unuse_commit_buffer(commit, buffer);
930 return saw_signature;
931}
932
933int remove_signature(struct strbuf *buf)
934{
935 const char *line = buf->buf;
936 const char *tail = buf->buf + buf->len;
937 int in_signature = 0;
938 const char *sig_start = NULL;
939 const char *sig_end = NULL;
940
941 while (line < tail) {
942 const char *next = memchr(line, '\n', tail - line);
943 next = next ? next + 1 : tail;
944
945 if (in_signature && line[0] == ' ')
946 sig_end = next;
947 else if (starts_with(line, gpg_sig_header) &&
948 line[gpg_sig_header_len] == ' ') {
949 sig_start = line;
950 sig_end = next;
951 in_signature = 1;
952 } else {
953 if (*line == '\n')
954 /* dump the whole remainder of the buffer */
955 next = tail;
956 in_signature = 0;
957 }
958 line = next;
959 }
960
961 if (sig_start)
962 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
963
964 return sig_start != NULL;
965}
966
967static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
968{
969 struct merge_remote_desc *desc;
970 struct commit_extra_header *mergetag;
971 char *buf;
972 unsigned long size, len;
973 enum object_type type;
974
975 desc = merge_remote_util(parent);
976 if (!desc || !desc->obj)
977 return;
978 buf = read_object_file(&desc->obj->oid, &type, &size);
979 if (!buf || type != OBJ_TAG)
980 goto free_return;
981 len = parse_signature(buf, size);
982 if (size == len)
983 goto free_return;
984 /*
985 * We could verify this signature and either omit the tag when
986 * it does not validate, but the integrator may not have the
987 * public key of the signer of the tag he is merging, while a
988 * later auditor may have it while auditing, so let's not run
989 * verify-signed-buffer here for now...
990 *
991 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
992 * warn("warning: signed tag unverified.");
993 */
994 mergetag = xcalloc(1, sizeof(*mergetag));
995 mergetag->key = xstrdup("mergetag");
996 mergetag->value = buf;
997 mergetag->len = size;
998
999 **tail = mergetag;
1000 *tail = &mergetag->next;
1001 return;
1002
1003free_return:
1004 free(buf);
1005}
1006
1007int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1008{
1009 struct strbuf payload = STRBUF_INIT;
1010 struct strbuf signature = STRBUF_INIT;
1011 int ret = 1;
1012
1013 sigc->result = 'N';
1014
1015 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1016 goto out;
1017 ret = check_signature(payload.buf, payload.len, signature.buf,
1018 signature.len, sigc);
1019
1020 out:
1021 strbuf_release(&payload);
1022 strbuf_release(&signature);
1023
1024 return ret;
1025}
1026
1027
1028
1029void append_merge_tag_headers(struct commit_list *parents,
1030 struct commit_extra_header ***tail)
1031{
1032 while (parents) {
1033 struct commit *parent = parents->item;
1034 handle_signed_tag(parent, tail);
1035 parents = parents->next;
1036 }
1037}
1038
1039static void add_extra_header(struct strbuf *buffer,
1040 struct commit_extra_header *extra)
1041{
1042 strbuf_addstr(buffer, extra->key);
1043 if (extra->len)
1044 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1045 else
1046 strbuf_addch(buffer, '\n');
1047}
1048
1049struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1050 const char **exclude)
1051{
1052 struct commit_extra_header *extra = NULL;
1053 unsigned long size;
1054 const char *buffer = get_commit_buffer(commit, &size);
1055 extra = read_commit_extra_header_lines(buffer, size, exclude);
1056 unuse_commit_buffer(commit, buffer);
1057 return extra;
1058}
1059
1060int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1061{
1062 struct commit_extra_header *extra, *to_free;
1063 int res = 0;
1064
1065 to_free = read_commit_extra_headers(commit, NULL);
1066 for (extra = to_free; !res && extra; extra = extra->next) {
1067 if (strcmp(extra->key, "mergetag"))
1068 continue; /* not a merge tag */
1069 res = fn(commit, extra, data);
1070 }
1071 free_commit_extra_headers(to_free);
1072 return res;
1073}
1074
1075static inline int standard_header_field(const char *field, size_t len)
1076{
1077 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1078 (len == 6 && !memcmp(field, "parent", 6)) ||
1079 (len == 6 && !memcmp(field, "author", 6)) ||
1080 (len == 9 && !memcmp(field, "committer", 9)) ||
1081 (len == 8 && !memcmp(field, "encoding", 8)));
1082}
1083
1084static int excluded_header_field(const char *field, size_t len, const char **exclude)
1085{
1086 if (!exclude)
1087 return 0;
1088
1089 while (*exclude) {
1090 size_t xlen = strlen(*exclude);
1091 if (len == xlen && !memcmp(field, *exclude, xlen))
1092 return 1;
1093 exclude++;
1094 }
1095 return 0;
1096}
1097
1098static struct commit_extra_header *read_commit_extra_header_lines(
1099 const char *buffer, size_t size,
1100 const char **exclude)
1101{
1102 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1103 const char *line, *next, *eof, *eob;
1104 struct strbuf buf = STRBUF_INIT;
1105
1106 for (line = buffer, eob = line + size;
1107 line < eob && *line != '\n';
1108 line = next) {
1109 next = memchr(line, '\n', eob - line);
1110 next = next ? next + 1 : eob;
1111 if (*line == ' ') {
1112 /* continuation */
1113 if (it)
1114 strbuf_add(&buf, line + 1, next - (line + 1));
1115 continue;
1116 }
1117 if (it)
1118 it->value = strbuf_detach(&buf, &it->len);
1119 strbuf_reset(&buf);
1120 it = NULL;
1121
1122 eof = memchr(line, ' ', next - line);
1123 if (!eof)
1124 eof = next;
1125 else if (standard_header_field(line, eof - line) ||
1126 excluded_header_field(line, eof - line, exclude))
1127 continue;
1128
1129 it = xcalloc(1, sizeof(*it));
1130 it->key = xmemdupz(line, eof-line);
1131 *tail = it;
1132 tail = &it->next;
1133 if (eof + 1 < next)
1134 strbuf_add(&buf, eof + 1, next - (eof + 1));
1135 }
1136 if (it)
1137 it->value = strbuf_detach(&buf, &it->len);
1138 return extra;
1139}
1140
1141void free_commit_extra_headers(struct commit_extra_header *extra)
1142{
1143 while (extra) {
1144 struct commit_extra_header *next = extra->next;
1145 free(extra->key);
1146 free(extra->value);
1147 free(extra);
1148 extra = next;
1149 }
1150}
1151
1152int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1153 struct commit_list *parents, struct object_id *ret,
1154 const char *author, const char *sign_commit)
1155{
1156 struct commit_extra_header *extra = NULL, **tail = &extra;
1157 int result;
1158
1159 append_merge_tag_headers(parents, &tail);
1160 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1161 author, sign_commit, extra);
1162 free_commit_extra_headers(extra);
1163 return result;
1164}
1165
1166static int find_invalid_utf8(const char *buf, int len)
1167{
1168 int offset = 0;
1169 static const unsigned int max_codepoint[] = {
1170 0x7f, 0x7ff, 0xffff, 0x10ffff
1171 };
1172
1173 while (len) {
1174 unsigned char c = *buf++;
1175 int bytes, bad_offset;
1176 unsigned int codepoint;
1177 unsigned int min_val, max_val;
1178
1179 len--;
1180 offset++;
1181
1182 /* Simple US-ASCII? No worries. */
1183 if (c < 0x80)
1184 continue;
1185
1186 bad_offset = offset-1;
1187
1188 /*
1189 * Count how many more high bits set: that's how
1190 * many more bytes this sequence should have.
1191 */
1192 bytes = 0;
1193 while (c & 0x40) {
1194 c <<= 1;
1195 bytes++;
1196 }
1197
1198 /*
1199 * Must be between 1 and 3 more bytes. Longer sequences result in
1200 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1201 */
1202 if (bytes < 1 || 3 < bytes)
1203 return bad_offset;
1204
1205 /* Do we *have* that many bytes? */
1206 if (len < bytes)
1207 return bad_offset;
1208
1209 /*
1210 * Place the encoded bits at the bottom of the value and compute the
1211 * valid range.
1212 */
1213 codepoint = (c & 0x7f) >> bytes;
1214 min_val = max_codepoint[bytes-1] + 1;
1215 max_val = max_codepoint[bytes];
1216
1217 offset += bytes;
1218 len -= bytes;
1219
1220 /* And verify that they are good continuation bytes */
1221 do {
1222 codepoint <<= 6;
1223 codepoint |= *buf & 0x3f;
1224 if ((*buf++ & 0xc0) != 0x80)
1225 return bad_offset;
1226 } while (--bytes);
1227
1228 /* Reject codepoints that are out of range for the sequence length. */
1229 if (codepoint < min_val || codepoint > max_val)
1230 return bad_offset;
1231 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1232 if ((codepoint & 0x1ff800) == 0xd800)
1233 return bad_offset;
1234 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1235 if ((codepoint & 0xfffe) == 0xfffe)
1236 return bad_offset;
1237 /* So are anything in the range U+FDD0..U+FDEF. */
1238 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1239 return bad_offset;
1240 }
1241 return -1;
1242}
1243
1244/*
1245 * This verifies that the buffer is in proper utf8 format.
1246 *
1247 * If it isn't, it assumes any non-utf8 characters are Latin1,
1248 * and does the conversion.
1249 */
1250static int verify_utf8(struct strbuf *buf)
1251{
1252 int ok = 1;
1253 long pos = 0;
1254
1255 for (;;) {
1256 int bad;
1257 unsigned char c;
1258 unsigned char replace[2];
1259
1260 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1261 if (bad < 0)
1262 return ok;
1263 pos += bad;
1264 ok = 0;
1265 c = buf->buf[pos];
1266 strbuf_remove(buf, pos, 1);
1267
1268 /* We know 'c' must be in the range 128-255 */
1269 replace[0] = 0xc0 + (c >> 6);
1270 replace[1] = 0x80 + (c & 0x3f);
1271 strbuf_insert(buf, pos, replace, 2);
1272 pos += 2;
1273 }
1274}
1275
1276static const char commit_utf8_warn[] =
1277N_("Warning: commit message did not conform to UTF-8.\n"
1278 "You may want to amend it after fixing the message, or set the config\n"
1279 "variable i18n.commitencoding to the encoding your project uses.\n");
1280
1281int commit_tree_extended(const char *msg, size_t msg_len,
1282 const struct object_id *tree,
1283 struct commit_list *parents, struct object_id *ret,
1284 const char *author, const char *sign_commit,
1285 struct commit_extra_header *extra)
1286{
1287 int result;
1288 int encoding_is_utf8;
1289 struct strbuf buffer;
1290
1291 assert_oid_type(tree, OBJ_TREE);
1292
1293 if (memchr(msg, '\0', msg_len))
1294 return error("a NUL byte in commit log message not allowed.");
1295
1296 /* Not having i18n.commitencoding is the same as having utf-8 */
1297 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1298
1299 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1300 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1301
1302 /*
1303 * NOTE! This ordering means that the same exact tree merged with a
1304 * different order of parents will be a _different_ changeset even
1305 * if everything else stays the same.
1306 */
1307 while (parents) {
1308 struct commit *parent = pop_commit(&parents);
1309 strbuf_addf(&buffer, "parent %s\n",
1310 oid_to_hex(&parent->object.oid));
1311 }
1312
1313 /* Person/date information */
1314 if (!author)
1315 author = git_author_info(IDENT_STRICT);
1316 strbuf_addf(&buffer, "author %s\n", author);
1317 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1318 if (!encoding_is_utf8)
1319 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1320
1321 while (extra) {
1322 add_extra_header(&buffer, extra);
1323 extra = extra->next;
1324 }
1325 strbuf_addch(&buffer, '\n');
1326
1327 /* And add the comment */
1328 strbuf_add(&buffer, msg, msg_len);
1329
1330 /* And check the encoding */
1331 if (encoding_is_utf8 && !verify_utf8(&buffer))
1332 fprintf(stderr, _(commit_utf8_warn));
1333
1334 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1335 result = -1;
1336 goto out;
1337 }
1338
1339 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1340out:
1341 strbuf_release(&buffer);
1342 return result;
1343}
1344
1345define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1346static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1347
1348struct merge_remote_desc *merge_remote_util(struct commit *commit)
1349{
1350 return *merge_desc_slab_at(&merge_desc_slab, commit);
1351}
1352
1353void set_merge_remote_desc(struct commit *commit,
1354 const char *name, struct object *obj)
1355{
1356 struct merge_remote_desc *desc;
1357 FLEX_ALLOC_STR(desc, name, name);
1358 desc->obj = obj;
1359 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1360}
1361
1362struct commit *get_merge_parent(const char *name)
1363{
1364 struct object *obj;
1365 struct commit *commit;
1366 struct object_id oid;
1367 if (get_oid(name, &oid))
1368 return NULL;
1369 obj = parse_object(the_repository, &oid);
1370 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1371 if (commit && !merge_remote_util(commit))
1372 set_merge_remote_desc(commit, name, obj);
1373 return commit;
1374}
1375
1376/*
1377 * Append a commit to the end of the commit_list.
1378 *
1379 * next starts by pointing to the variable that holds the head of an
1380 * empty commit_list, and is updated to point to the "next" field of
1381 * the last item on the list as new commits are appended.
1382 *
1383 * Usage example:
1384 *
1385 * struct commit_list *list;
1386 * struct commit_list **next = &list;
1387 *
1388 * next = commit_list_append(c1, next);
1389 * next = commit_list_append(c2, next);
1390 * assert(commit_list_count(list) == 2);
1391 * return list;
1392 */
1393struct commit_list **commit_list_append(struct commit *commit,
1394 struct commit_list **next)
1395{
1396 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1397 new_commit->item = commit;
1398 *next = new_commit;
1399 new_commit->next = NULL;
1400 return &new_commit->next;
1401}
1402
1403const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1404{
1405 int key_len = strlen(key);
1406 const char *line = msg;
1407
1408 while (line) {
1409 const char *eol = strchrnul(line, '\n');
1410
1411 if (line == eol)
1412 return NULL;
1413
1414 if (eol - line > key_len &&
1415 !strncmp(line, key, key_len) &&
1416 line[key_len] == ' ') {
1417 *out_len = eol - line - key_len - 1;
1418 return line + key_len + 1;
1419 }
1420 line = *eol ? eol + 1 : NULL;
1421 }
1422 return NULL;
1423}
1424
1425/*
1426 * Inspect the given string and determine the true "end" of the log message, in
1427 * order to find where to put a new Signed-off-by: line. Ignored are
1428 * trailing comment lines and blank lines. To support "git commit -s
1429 * --amend" on an existing commit, we also ignore "Conflicts:". To
1430 * support "git commit -v", we truncate at cut lines.
1431 *
1432 * Returns the number of bytes from the tail to ignore, to be fed as
1433 * the second parameter to append_signoff().
1434 */
1435size_t ignore_non_trailer(const char *buf, size_t len)
1436{
1437 size_t boc = 0;
1438 size_t bol = 0;
1439 int in_old_conflicts_block = 0;
1440 size_t cutoff = wt_status_locate_end(buf, len);
1441
1442 while (bol < cutoff) {
1443 const char *next_line = memchr(buf + bol, '\n', len - bol);
1444
1445 if (!next_line)
1446 next_line = buf + len;
1447 else
1448 next_line++;
1449
1450 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1451 /* is this the first of the run of comments? */
1452 if (!boc)
1453 boc = bol;
1454 /* otherwise, it is just continuing */
1455 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1456 in_old_conflicts_block = 1;
1457 if (!boc)
1458 boc = bol;
1459 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1460 ; /* a pathname in the conflicts block */
1461 } else if (boc) {
1462 /* the previous was not trailing comment */
1463 boc = 0;
1464 in_old_conflicts_block = 0;
1465 }
1466 bol = next_line - buf;
1467 }
1468 return boc ? len - boc : len - cutoff;
1469}